Recent advances in Convolutional Neural Network (CNN) model interpretability have led to impressive progress in visualizing and understanding model predictions. In particular, gradient-based visual attention methods have driven much recent effort in using visual attention maps as a means for visual explanations. A key problem, however, is these methods are designed for classification and categorization tasks, and their extension to explaining generative models, e.g., variational autoencoders (VAE) is not trivial. In this work, we take a step towards bridging this crucial gap, proposing the first technique to visually explain VAEs by means of gradient-based attention. We present methods to generate visual attention from the learned latent space, and also demonstrate such attention explanations serve more than just explaining VAE predictions. We show how these attention maps can be used to localize anomalies in images, demonstrating state-of-the-art performance on the MVTec-AD dataset. We also show how they can be infused into model training, helping bootstrap the VAE into learning improved latent space disentanglement, demonstrated on the Dsprites dataset.

No abstract

No abstract

Person identification across nonoverlapping cameras, also known as person reidentification, aims to match people at different times and locations. Reidentifying people is of great importance in crucial applications such as widearea surveillance and visual tracking. Due to the appearance variations in pose, illumination, and occlusion in different camera views, person reidentification is inherently difficult. To address these challenges, a reference-based method is proposed for person reidentification across different cameras. Instead of directly matching people by their appearance, the matching is conducted in a reference space where the descriptor for a person is translated from the original color or texture descriptors to similarity measures between this person and the exemplars in the reference set. A subspace is first learned in which the correlations of the reference data from different cameras are maximized using regularized canonical correlation analysis (RCCA). For reidentification, the gallery data and the probe data are projected onto this RCCA subspace and the reference descriptors (RDs) of the gallery and probe are generated by computing the similarity between them and the reference data. The identity of a probe is determined by comparing the RD of the probe and the RDs of the gallery. A reranking step is added to further improve the results using a saliency-based matching scheme. Experiments on publicly available datasets show that the proposed method outperforms most of the state-of-the-art approaches.

Tracking multiple targets in nonoverlapping cameras are challenging since the observations of the same targets are often separated by time and space. There might be significant appearance change of a target across camera views caused by variations in illumination conditions, poses, and camera imaging characteristics. Consequently, the same target may appear very different in two cameras. Therefore, associating tracks in different camera views directly based on their appearance similarity is difficult and prone to error. In most previous methods, the appearance similarity is computed either using color histograms or based on pretrained brightness transfer function that maps color between cameras. In this paper, a novel reference set based appearance model is proposed to improve multitarget tracking in a network of nonoverlapping cameras. Contrary to previous work, a reference set is constructed for a pair of cameras, containing subjects appearing in both camera views. For track association, instead of directly comparing the appearance of two targets in different camera views, they are compared indirectly via the reference set. Besides global color histograms, texture and shape features are extracted at different locations of a target, and AdaBoost is used to learn the discriminative power of each feature. The effectiveness of the proposed method over the state of the art on two challenging real-world multicamera video data sets is demonstrated by thorough experiments.